Investor-Sentiment-and-the-Cross-Section-of-Stock-

THEJOURNALOFFINANCE•VOL.LXI,NO.4•AUGUST2006InvestorSentimentandtheCross-SectionofStockReturnsMALCOLMBAKERandJEFFREYWURGLER∗ABSTRACTWestudyhowinvestorsentimentaffectsthecross-sectionofstockreturns.Wepre-dictthatawaveofinvestorsentimenthaslargereffectsonsecuritieswhosevalua-tionsarehighlysubjectiveanddifficulttoarbitrage.Consistentwiththisprediction,wefindthatwhenbeginning-of-periodproxiesforsentimentarelow,subsequentre-turnsarerelativelyhighforsmallstocks,youngstocks,highvolatilitystocks,un-profitablestocks,non-dividend-payingstocks,extremegrowthstocks,anddistressedstocks.Whensentimentishigh,ontheotherhand,thesecategoriesofstockearnrelativelylowsubsequentreturns.CLASSICALFINANCETHEORYLEAVESNOROLEFORINVESTORSENTIMENT.Rather,thistheoryarguesthatcompetitionamongrationalinvestors,whodiversifytoopti-mizethestatisticalpropertiesoftheirportfolios,willleadtoanequilibriuminwhichpricesequaltherationallydiscountedvalueofexpectedcashflows,andinwhichthecross-sectionofexpectedreturnsdependsonlyonthecross-sectionofsystematicrisks.1Evenifsomeinvestorsareirrational,classicaltheoryar-gues,theirdemandsareoffsetbyarbitrageursandthushavenosignificantimpactonprices.Inthispaper,wepresentevidencethatinvestorsentimentmayhavesignifi-canteffectsonthecross-sectionofstockprices.Westartwithsimpletheoreticalpredictions.Becauseamispricingistheresultofanuninformeddemandshockinthepresenceofabindingarbitrageconstraint,wepredictthatabroad-basedwaveofsentimenthascross-sectionaleffects(thatis,doesnotsimplyraiseorlowerallpricesequally)whensentiment-baseddemandsorarbitrage∗BakerisattheHarvardBusinessSchoolandNationalBureauofEconomicResearch;WurglerisattheNYUSternSchoolofBusinessandtheNationalBureauofEconomicResearch.Wethankananonymousreferee,RobStambaugh(theeditor),NedElton,WayneFerson,XavierGabaix,MartyGruber,LisaKramer,OwenLamont,MartinLettau,AnthonyLynch,JayShanken,MeirStatman,SheridanTitman,andJeremySteinforhelpfulcomments,aswellasparticipantsofconferencesorseminarsatBaruchCollege,BostonCollege,theChicagoQuantitativeAlliance,EmoryUniversity,theFederalReserveBankofNewYork,HarvardUniversity,IndianaUniversity,MichiganStateUniversity,NBER,theNorwegianSchoolofEconomicsandBusiness,NorwegianSchoolofManagement,NewYorkUniversity,StockholmSchoolofEconomics,TulaneUniversity,theUniversityofAmsterdam,theUniversityofBritishColumbia,theUniversityofIllinois,theUniversityofKentucky,theUniversityofMichigan,theUniversityofNotreDame,theUniversityofTexas,andtheUniversityofWisconsin.WegratefullyacknowledgefinancialsupportfromtheQGroupandtheDivisionofResearchoftheHarvardBusinessSchool.1SeeGomes,Kogan,andZhang(2003)forarecentmodelinthistradition.16451646TheJournalofFinanceconstraintsvaryacrossstocks.Inpractice,thesetwodistinctchannelsleadtoquitesimilarpredictionsbecausestocksthatarelikelytobemostsensitivetospeculativedemand,thosewithhighlysubjectivevaluations,alsotendtobetheriskiestandcostliesttoarbitrage.Concretely,then,theorysuggeststwodistinctchannelsthroughwhichthesharesofcertainfirms—newer,smaller,morevolatile,unprofitable,non-dividendpaying,distressedorwithextremegrowthpotential,andfirmswithanalogouscharacteristics—arelikelytobemoreaffectedbyshiftsininvestorsentiment.Toinvestigatethispredictionempirically,andtogetamoretangiblesenseoftheintrinsicallyelusiveconceptofinvestorsentiment,westartwithasummaryoftherisesandfallsinU.S.marketsentimentfrom1961throughtheInternetbubble.Thissummaryisbasedonanecdotalaccountsandthusbyitsnaturecanonlybeasuggestive,expostcharacterizationoffluctuationsinsentiment.Nonetheless,itsbasicmessageappearsbroadlyconsistentwithourtheoreticalpredictionsandsuggeststhatmorerigoroustestsarewarranted.Ourmainempiricalapproachisasfollows.Becausecross-sectionalpatternsofsentiment-drivenmispricingwouldbedifficulttoidentifydirectly,weex-aminewhethercross-sectionalpredictabilitypatternsinstockreturnsdependuponproxiesforbeginning-of-periodsentiment.Forexample,lowfuturereturnsonyoungfirmsrelativetooldfirms,conditionalonhighvaluesforproxiesforbeginning-of-periodsentiment,wouldbeconsistentwiththeexanterelativeovervaluationofyoungfirms.Asusual,wearemindfulofthejointhypothesisproblemthatanypredictabilitypatternswefindactuallyreflectcompensationforsystematicrisks.Thefirststepistogatherproxiesforinvestorsentimentthatwecanuseastime-seriesconditioningvariables.Sincetherearenoperfectand/oruncontro-versialproxiesforinvestorsentiment,ourapproachisnecessarilypractical.Specifically,weconsideranumberofproxiessuggestedinrecentworkandformacompositesentimentindexbasedontheirfirstprincipalcomponent.Toreducethelikelihoodthattheseproxiesareconnectedtosystematicrisk,wealsoformanindexbasedonsentimentproxiesthathavebeenorthogonalizedtoseveralmacroeconomicconditions.Thesentimentindexesvisiblylineupwithhistoricalaccountsofbubblesandcrashes.Wethentesthowthecross-sectionofsubsequentstockreturnsvarieswithbeginning-of-periodsentiment.Usingmonthlystockreturnsbetween1963and2001,westartbyformingequal-weighteddecileportfoliosbasedonseveralfirmcharacteristics.(Ourtheorypredicts,andtheemp